Language Control and Intra-Sentential Codeswitching among Bilingual Children with and without Developmental Language Disorder
Round 1
Reviewer 1 Report
In the manuscript, “Language control and intra-sentential codeswitching among bilingual children with and without Developmental Language Disorder,” the authors examined children’s productions in a sentence repetition task in Hebrew and English that included sentences with varying amounts of code-switching. The authors systematically manipulated the location and duration of the code-switch within the prepositional phrase. The key measure of interest was cross-language errors, or non-target CS, in which children produced words in the opposite language from the model sentence. Overall, children produced more non-target CS during sentences that already contained code-switching compared to single-language sentences. In particular, children produced more cross-language errors during the Hebrew sentences that contained switches into English, especially when the full prepositional phrase was presented in English. Most cross-language errors occurred within the temporal phrase following the prepositional phrase. However, children with DLD showed some different patterns in that they produced more cross-language errors overall, showed less of a directionality effect, and were less likely than TLD peers to produce cross-language errors on determiners only.
This project makes an important contribution to the study of language control in bilingual children with DLD. With its novel application of a sentence repetition task, commonly used in language assessment, it contributes to the effort to identify differences that may help to distinguish children with DLD from typically developing peers. However, I have several questions related to the methodology; addressing these may make it easier to interpret the findings.
1. Language cut scores: Although consistent with past literature, why is it that a lower cut-off is used for Hebrew than for English? Is it related to the nature of the test or the expected Hebrew skills of the children? How were the different cut points based on length of Hebrew exposure determined? Does amount of Hebrew exposure play any role, or only the duration?
2. Relative dominance: There were no significant differences between Hebrew vs. English scores for each group as a whole, but did individual children differ in their level of Hebrew vs. English dominance? Is it possible that differences in directionality effects for the TLD and DLD groups could reflect differences in relative dominance? Can the English and Hebrew scores be compared in a meaningful way, given that they come from different tests normed on different samples? Is there any other way to quantify relative dominance?
3. Exposure: The authors report the duration of Hebrew exposure. Is there any information about current proportion of exposure to English vs. Hebrew that could contribute to directionality effects?
4. CS sentences: How did the different CS conditions compare to typical patterns for Hebrew-English code-switching? Do speakers in this community code-switch frequently? Do switches tend to be more single-word insertions or full phrases? How might the naturalness or acceptability of the different code-switching variations affect repetition performance?
5. CS experience: Were any data collected on children’s exposure to or experience with code-switching? CS experience might affect their repetition of the CS sentences and could be a point to consider in the limitations and future directions section. The possibility of more CS experience in children with DLD is raised briefly in the Discussion but without any concrete suggestion to collect such data.
6. SRep task: What instructions were children given? Did the instructions mention anything about language or the possibility that the sentences to be repeated would contain both languages?
7. Defining cross-language errors: Did children ever fail to repeat the code-switched words in the other language (e.g., produce the whole sentence in Hebrew even when parts of the PP were presented in English)? Were these also counted as cross-language errors? The authors mention that CS within the PP was not examined in the current study because it was already examined elsewhere. However, if the target sentence included a switch on only the noun but the child switched for the whole PP (or vice versa), it still seems like this would be important to include when comparing the location of cross-language intrusions.
8. Outcome variables: I found the count values for the outcome variable a little confusing to interpret given the different number of children in each group and the different number of unilingual and CS trials. I wonder if a proportion of items containing non-target CS in each condition and group, averaged across participants, might be easier to interpret and compare across trial types and groups. For example, in Figure 2, I was not sure what the values represented and how much variability there was within groups. Similarly, in Table 4, a raw count of instances within each group does not give any indication of how many children contributed those instances – were these few instances from several children or many instances from only a few children?
9. Switch location: In Tables 5 and 6, are these data collapsed across sentence types? Did the unilingual sentences differ from the CS sentences in the location of non-target CS?
1. Analysis approach: The authors used a Friedman test, which I understand to be a non-parametric equivalent of a one-way ANOVA with repeated measures. Is there a way to also examine the effect of group and potential interactions between effects of condition and group within a single analysis, instead of running multiple analyses for the whole sample and then subgroups? Would it be appropriate to consider a mixed-effects model including condition, group, their interaction, and a random intercept to account for repeated measures within subjects?
1. In the discussion of directionality, how might the findings in the current study be affected by the unique aspects of a sentence repetition task, compared to a spontaneous production task?
1. In the discussion of differences for children with DLD, I am still not quite clear on why they might show more Det+N switches rather than determiner only. Could this phenomenon be specific to temporal phrases, which could lend themselves to be learned as chunks more than other types of Det+N phrases?
In addition to these points about methodology, one question regarding theoretical framing is the statement in the abstract and elsewhere that code-switches in this paradigm reflect difficulty in language control. Is it possible that at least some such switches reflect an effort to fill an expressive gap rather than a lack of control? The distinction between cross-language errors and intra-sentential CS as a resourceful mode of expression could be clearer.
A few final, smaller points:
1. The authors frequently provide direct quotes from authors. Sometimes this can be beneficial for clarity, but I wonder if some of these could be paraphrased.
2. Organization of sections: The headings for section 1 and section 2 of the introduction might be clarified. The first section also discusses language control in bilingual children, and the second section does not only discuss children. Maybe “language control and domain-general cognitive control” would be a more accurate heading for the second section?
3. Line 178: In reference to the work by Gross & Kaushanskaya (2022) in the introduction, the authors use the term “cross-language switch.” Cross-speaker switch would be more consistent with other work, such as by Raichlin et al., and more directly pinpoints the key feature of a language switch across rather than within speakers.
4. The authors cite Kootstra et al., 2010 as a basis for their hypothesis (line 314). It might be beneficial to explain more about this study within the Introduction, as it is another example of manipulating the structure of CS in a model sentence, even though it is not a sentence repetition task.
6. Lines 861-864: I had some difficulty following the explanations in this paragraph, especially the point presented here about being “a function of the nature of bilingual DLD, viz., a language impairment in both languages.” The final sentence in this paragraph seems to be more directly related to this point than the sentences that immediately follow it.
Thank you for the opportunity to review this manuscript.
Author Response
Please see attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper aims at evaluating language control and intra-sentential codeswitching through a sentence repetition task in bilingual children with typical language development or developmental language disorder. The main findings show that codeswitching is more frequent in bilingual than in monolingual stimuli and in children with DLD than in children with TLD. Data also revealed a directionality effect for children with TLD, as well as more frequent codeswitches for final temporal phrases.
This study addresses interesting questions about language control in bilingual children, and specifically in bilingual children with DLD, a population which is not often considered. The experimental task is rarely used in studies of bilingualism and is particularly relevant for studying DLD. The connections with the literature are rich and relevant. The paper is globally clear and well-constructed.
General comments:
1. Given the small number of DLD participants and the frequent heterogeneity of performance in these children, I have some concerns about the impact of possible individual differences on the results (the high number of codeswitches in the DLD group might be attributed to one or two children specifically). It would be interesting to have some information about this aspect of the data, and it would also reinforce the results and interpretations.
2. Qualitative differences that are sometimes not supported by statistically significative results are largely reported and don’t seem really useful to me. See for example page 18, line 633 to 644. These non-significative results should not be used in the discussion and conclusions. For example, page 18 line 66 and again page 24 line 915, when stating that codeswitches involved more often determiners when it was not statistically supported, or page 24 line 931, when stating that both groups had more intrusions in the temporal phrase when it was not the case for children with DLD.
3. Language measures which were used to establish the group repartition could have been used to investigate proficiency and its impact on codeswitches and on the directionality effect observed in TLD children or on the lack of directionality effect observed in DLD children. I would like the authors to consider this analysis or at least to explain why they did not do it.
4. It might be interesting to report effect sizes throughout the methods and results.
Specific comments:
Introduction
1. Page 3: When reporting results (Gutiérrez et al., Iluz-Cohen & Walters, etc.), please specify the status of each language (home vs. societal, dominant vs. non-dominant, etc.) as it was made when citing Smolak’s study for example, for the reader to better understand the results and interpretations regarding the influence of linguistic and socio-pragmatic skills. The distinction between exposure and proficiency was also not always clear. The systematic use of specific words such as home vs. societal for exposure and dominant vs. non-dominant for proficiency could facilitate the reading.
2. Page 5: Same comment for the study of Gross & Kaushanskaya, it would be helpful to specify the status of each language.
3. Page 7, line 321-322: Same comment, to facilitate the understanding of the study, please specify which is the minority language and which is the majority one. I also have some concerns about the fact that the home language is considered as the minority one and the societal language is considered as the majority one. Reporting the time of exposure per day or per week could have supported this statement.
Methods
1. Page 7, line 358: Please define sequential bilingualism, as it could slightly differ from one author to another.
2. Page 8, line 369: Il don’t understand why the cutoff scores are different for the English and Hebrew tasks.
3. Page 8, line 376: Why did you use local norms and not the norms used to establish the groups?
4. Page 8, line 387: How did you choose the cutoff scores to establish the categories?
5. Table 1: I have some concerns about DLD children being older than TLD children… It is not discussed in the paper but it might be necessary to do so. I am also surprised that some TLD children were recruited in language classes, while DLD children were identified as so only through the evaluation performed for the study (and not through a prior diagnosis for example). This might weaken the group distinction and might require some discussions. Finally, the results of the NVIQ could be reported in the table, to attest for an absence of difference between the two groups.
6. Page 12, line 487: The codeswitches do not appear in bold in the stimulus sentences.
Results
1. Table 2: Providing the percentages would be helpful, given the large discrepancy of effectives.
Discussion
1. Page 19, line 677: In Hebrew but not in English, the statement should be nuanced.
2. Page 20, line 737: For TLD children but nor for DLD children, please specify.
3. Page 20, line 756-763: This section could be improved, in relation with the literature, by highlighting possible differences with former studies for example or developing some interpretations supported by the findings and other experiments.
4. Page 22, line 836-838: The formulation was not very clear to me and made the experiment difficult to understand.
5. Page 23, line 886: It seems to me that it was the case for the TLD children only, this should be specified.
Author Response
Please see attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
I am very satisfied with the responses and the modifications which were made by the authors. The result section is now much more reliable while improving also the discussion.
